AI Health

Friday Roundup

The AI Health Friday Roundup highlights the week’s news and publications related to artificial intelligence, data science, public health, and clinical research.

February 25, 2022

In today’s Roundup: when docs spread misinformation; deep learning holds reins in fusion reactor; parsing NC regulations on syringes; sudden collapse of pain clinics leaves patients stranded; trade secrets, patents and bioscience; remembering Paul Farmer; the “wicked problem” posed by retracted scientific papers; fighting ageism in AI; groundswell gathers for federal privacy protections; much more:

Deep Breaths

In a timely intervention, computer scientist David Andersen offers a Twitter thread reassuring anxious fans that the popular word game Wordle – recently purchased by the New York Times – has not, contrary to online rumors, suddenly gotten harder.

AI, STATISTICS & DATA SCIENCE

“A core challenge is to shape and maintain a high-temperature plasma within the tokamak vessel. This requires high-dimensional, high-frequency, closed-loop control using magnetic actuator coils, further complicated by the diverse requirements across a wide range of plasma configurations.” A paper published in Nature by Degrave and colleagues introduces an application for machine learning somewhat outside of our usual Roundup ambit: a reinforcement-learning system for keeping solar-heat plasmas bottled up inside of tokamak-design fusion reactors.
“Cohen and Cope were among a handful of eccentrics pushing computers to go against their nature as cold, calculating things. The still-nascent field of AI had its focus set squarely on solid concepts like reasoning and planning, or on tasks like playing chess and checkers or solving mathematical problems. Most AI researchers balked at the notion of creative machines.” In an article in Science News, Richard Moss explores the intersection between artificial intelligence and creative and artistic endeavors.
“…AI technologies may perpetuate existing ageism in society and undermine the quality of health and social care that older people receive. The data used by AI can be unrepresentative of older people or skewed by past ageist stereotypes, prejudice or discrimination. Flawed assumptions of how older people wish to live or interact with technology in their daily lives can also limit the design and reach of these technologies, and the way AI technologies are used can reduce intergenerational contact or deepen existing barriers to digital access.” The World Health Organization this month announced the publication of a new policy brief aimed at countering ageism in health AI applications.
“This all started when an eagle-eyed editor at SAGE Publishing noticed that 2 different referees had left identical comments on 2 different peer-reviews. That seemed like a sure-sign that someone was attempting to game our peer-review system and it gave us the idea to survey our peer-review comments for more cases like this.” A conversation between Retraction Watch and Sage Publishing’s Adam Day digs deep into a method that identifies bogus “paper mill” articles via recycled reviewer comments – but the process can be a complex one.
“Within the standard model of deep learning, we classify all possible fundamental building blocks of attention in terms of their source, target, and computational mechanism. We identify and study three most important mechanisms: additive activation attention, multiplicative output attention (output gating), and multiplicative synaptic attention (synaptic gating).” A preprint by Baldi and Vershynin available from arXiv reviews the roles played by different kinds of “attention” in the context of deep learning systems.
“To build the new model, the team at the University of Illinois took an abundance of findings from various fields and wove them together.” Quanta’s Yasemin Saplakoglu reports on work, recently published in Cell by Thornburg and colleagues, that demonstrates a new threshold in computer simulations of living cells.
“…all sorts of businesses — from home loan providers and banks to job recruitment services — use algorithmic systems to make automated decisions. In an effort to enable more oversight and control of technologies that make discriminatory decisions or create safety risks or other harms, the bill would require companies deploying automated systems to assess them, mitigate negative impacts and submit annual reports about those assessments to the FTC.” An article at Protocol by Kate Kaye focuses on a particular facet of the recently announced Algorithmic Accountability Act of 2022 – provisions that would require audits of AI systems involved in lending and employment decision-making.

BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH

“Although state law makes syringes available over-the-counter for everyone, there’s nothing in the law that prohibits a pharmacy from setting a stricter policy. NC Health News made calls to various pharmacies — chain and independent — across the state and found that the policies vary widely by store, and sometimes prove oddly difficult to find out.” In a story for North Carolina Health News, Clarissa Donnelly-DeRoven scrutinizes the prevailing confusion about who in NC is able to buy syringes, and what the law actually states.
“In a matter of days, Lags Medical, a sprawling network of privately owned pain clinics serving more than 20,000 patients throughout the state’s Central Valley and Central Coast, would shut its doors. Its patients, most of them working-class people reliant on government-funded insurance, were left without ready access to their medical records or handoffs to other physicians.” A story at Kaiser Health News by Anna Maria Barry-Jester and Jenny Gold explores the sudden collapse of network of pain clinics – and the thousands of patients left in the lurch.
In a research article published last week in JAMA, Simon and colleagues report results from a pragmatic randomized trial that found that a “low-intensity” online behavioral therapy skills training intervention for persons expressing suicidal ideation was associated with a significant increase in the risk of self-harm.
“Gov. Greg Abbott told state health agencies in Texas on Tuesday that medical treatments provided to transgender adolescents, widely considered to be the standard of care in medicine, should be classified as ‘child abuse’ under existing state law.” The New York Times’ Azeen Gharayshi reports on a recent, highly controversial directive from Texas Governor Greg Abbott that equates gender-affirming treatment for transgender children with abuse.
“…Today, petrochemical production spews out nearly 2% of the world’s greenhouse gas emissions. Now, researchers have taken an important step to vastly reduce that footprint, by using bacteria and waste gases from steel plants, rather than petroleum, as the starting ingredient for dozens of commodity chemicals.” At Science, Robert F. Service reports on a new, potentially more environmentally friendly approach to using microbes to process waste gases into industrial chemicals.

COMMUNICATIONS & DIGITAL SOCIETY

“The power of social media amplifies the message of the relatively few physicians making false claims, Rachel Moran, PhD, a postdoctoral scholar at the University of Washington’s Center for an Informed Public who studies the spread of misinformation and disinformation, said in an interview. This amplification ‘makes it seem like there is more of a split within the medical community,’ she said, noting that ‘misinformation thrives in this uncertainty.’” In a news article at JAMA, Rita Rubin explores the contours of the phenomenon of medical misinformation when the person creating or spreading it is a licensed physician.
“Addressing wicked problems involves bringing the disparate body of people and organizations affected by the problem together to build the will to shift perceptions of the problem. It also involves clarifying incentives for stakeholders to work with change processes, as their perceptions of wicked problems shift with re-evaluations of risk, changing values, and in response to emergent policy and technical resolution processes.” A paper by Wood and colleagues that reports on a case study conducted by the RISRS project (Reducing the Inadvertent Spread of Retracted Science) tackles the “wicked problem” of how to deal with published science that remains in circulation (and citation) even after it has been retracted.
“Even now, the field [of global health] often views its work from a perspective of constraint. With the limited resources we have, what is the most good we can do?…Paul hated that question. He preferred to flip it on its head: Given all the good we can do for our fellow humans, what are the resources we need to make it happen? He was unconstrained by small thinking.” At The Atlantic, Brown University physician and researcher Ashish K. Jha offers a moving eulogy and appreciation for physician and global health expert Paul Farmer, whose unexpected death this week sent shock waves through the medical and global health communities.
“The popular academic social networking site ResearchGate is appealing a 31 January ruling by a district court in Munich, Germany, which ruled that the site is responsible for scholarly papers infringing copyright that were uploaded to its website.” Chemistry World’s Dalmeet Singh Chawla reports on a recent German court ruling that found ResearchGate, a popular portal for accessing scientific papers while avoiding paywalls, is responsible for pirated material available on the site.
“The truth is that what counts as an ‘ideal’ applicant is a moving target. The process can be unfair—even unjust. Practical considerations also play a role, starting with the reality that programs have limited funding and advising capacity. Either way, the reasons for rejection are often as much about the professors and programs as they are about the applicants.” In a commentary at Science, University of Southern California Associate Dean Julie R. Posselt unpacks the complexities of the graduate admissions process.

POLICY

“For most biological innovations, patents and trade secrets are used together to protect intellectual property. One of the key reasons why both are used together is because patents only require a limited disclosure of information. It is common for life sciences companies to keep the ‘best mode’ for production of a product a trade secret, but file a patent for the end product with only the ‘sufficient’ disclosure to meet the bar of patentability.” A post by Matt Bauer at the Petrie-Flom Center’s Bill of Health blog tackles the thorny issue of intellectual property, patent protections, and trade secrets in the biomedical sciences.
“Health tech companies worried that an emerging patchwork of state privacy laws will drive up regulatory costs are joining privacy hardliners in the call for one nationwide standard to govern how they handle patient data.” A STAT News article (subscription required) by Mohana Ravindranath surveys the growing unease being felt by the health technology industry as individual US states, in the absence of an overarching federal framework, seek to develop their own privacy regulations for customer/patient data.
“…Americans are used to having privacy managed through a patchwork of consents, cookies and compliance, but that is insufficient. We need to move toward a culture that addresses and respects the underlying concepts of data security and privacy through federal legislation that sets a firm standard for data collection, retention and transfer.” On a similar note: an opinion piece at Protocol by Tatyana Bolton, Chris Riley, and Brandon Pugh advocates for swift action on a federal-level set of internet privacy protections comparable to the European Union’s General Data Protection Regulation (GDPR).
“Under the new policy, which will go into effect on 25 January, all NIH grant applications for projects that collect scientific data must include a ‘data management and sharing’ (DMS) plan that contains details about the software or tools needed to analyse the data, when and where the raw data will be published and any special considerations for accessing or distributing that data.” In a news feature in Nature, Max Kozlov examines the implications of the NIH’s new data-sharing policy, slated to go into effect early next year, and some potential downstream complications for scientists and researchers.